cryptodev HW (aesni) vs software

[Ivan Klymenko]

On Fri, 20 Nov 2015 19:29:20 +0000
RW via freebsd-questions <freebsd-questions at freebsd.org> wrote:

> On Fri, 20 Nov 2015 20:03:25 +0200
> Ivan Klymenko wrote:
> 
> > Hello everybody.
> > What's happening?
> > 
> > root at nonamehost:/ # openssl speed -evp aes-128-cbc -elapsed -engine
> > cryptodev engine "cryptodev" set.  
> 
> I'm not really sure what the actual question was,  but IIRC you
> shouldn't load cryptodev if the CPU has AES-NI because it gets used by
> default and it's faster to allow openssl to use AES-NI instructions
> directly.

You may not see the problem.

There are two identical server hardware

1) FreeBSD and cryptodev/aesni:

openssl speed -evp aes-128-cbc -elapsed
You have chosen to measure elapsed time instead of user CPU time.
Doing aes-128-cbc for 3s on 16 size blocks: 1546914 aes-128-cbc's in 3.05s
Doing aes-128-cbc for 3s on 64 size blocks: 1442599 aes-128-cbc's in 3.01s
Doing aes-128-cbc for 3s on 256 size blocks: 1230228 aes-128-cbc's in 3.05s
Doing aes-128-cbc for 3s on 1024 size blocks: 783365 aes-128-cbc's in 3.00s
Doing aes-128-cbc for 3s on 8192 size blocks: 184665 aes-128-cbc's in 3.03s
OpenSSL 1.0.1p-freebsd 9 Jul 2015
built on: date not available
options:bn(64,64) rc4(16x,int) des(idx,cisc,16,int) aes(partial) idea(int) blowfish(idx) 
compiler: clang
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
aes-128-cbc       8123.28k    30695.51k   103364.39k   267388.59k   499060.02k

2) Ubuntu default (aesni module is load):

openssl speed -evp aes-128-cbc -elapsed
You have chosen to measure elapsed time instead of user CPU time.
Doing aes-128-cbc for 3s on 16 size blocks: 122663419 aes-128-cbc's in 3.00s
Doing aes-128-cbc for 3s on 64 size blocks: 32869514 aes-128-cbc's in 3.00s
Doing aes-128-cbc for 3s on 256 size blocks: 8368810 aes-128-cbc's in 3.00s
Doing aes-128-cbc for 3s on 1024 size blocks: 2101689 
aes-128-cbc's in 3.00s
Doing aes-128-cbc for 3s on 8192 size blocks: 263094 aes-128-cbc's in 3.00s
OpenSSL 1.0.1f 6 Jan 2014
built on: Thu Jun 11 15:28:12 UTC 2015
options:bn(64,64) rc4(16x,int) des(idx,cisc,16,int) aes(partial) blowfish(idx) 
compiler: cc -fPIC -DOPENSSL_PIC -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -m64 -DL_ENDIAN -DTERMIO -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security -D_FORTIFY_SOURCE=2 -Wl,-Bsymbolic-functions -Wl,-z,relro -Wa,--noexecstack -Wall -DMD32_REG_T=int -DOPENSSL_IA32_SSE2 -DOPENSSL_BN_ASM_MONT -DOPENSSL_BN_ASM_MONT5 -DOPENSSL_BN_ASM_GF2m -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DMD5_ASM -DAES_ASM -DVPAES_ASM -DBSAES_ASM -DWHIRLPOOL_ASM -DGHASH_ASM
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
aes-128-cbc     654204.90k   701216.30k   714138.45k   717376.51k   718422.02k

3)FreeBSD and only aesni:

openssl speed -evp aes-128-cbc -elapsed
You have chosen to measure elapsed time instead of user CPU time.
Doing aes-128-cbc for 3s on 16 size blocks: 98895809 aes-128-cbc's in 3.01s
Doing aes-128-cbc for 3s on 64 size blocks: 26629435 aes-128-cbc's in 3.01s
Doing aes-128-cbc for 3s on 256 size blocks: 6850987 aes-128-cbc's in 3.02s
Doing aes-128-cbc for 3s on 1024 size blocks: 1719734 aes-128-cbc's in 3.02s
Doing aes-128-cbc for 3s on 8192 size blocks: 215447 aes-128-cbc's in 3.02s
OpenSSL 1.0.1p-freebsd 9 Jul 2015
built on: date not available
options:bn(64,64) rc4(16x,int) des(idx,cisc,16,int) aes(partial) idea(int) blowfish(idx) 
compiler: clang
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
aes-128-cbc     526074.33k   566619.04k   580085.64k   583961.07k   583753.37k

I hope the problem has now become more visible?

Thanks!